xref: /freebsd/sys/contrib/openzfs/module/zfs/bpobj.c (revision a90b9d0159070121c221b966469c3e36d912bf82)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or https://opensource.org/licenses/CDDL-1.0.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23  * Copyright (c) 2011, 2018 by Delphix. All rights reserved.
24  * Copyright (c) 2017 Datto Inc.
25  */
26 
27 #include <sys/bpobj.h>
28 #include <sys/zfs_context.h>
29 #include <sys/zfs_refcount.h>
30 #include <sys/dsl_pool.h>
31 #include <sys/zfeature.h>
32 #include <sys/zap.h>
33 
34 /*
35  * Return an empty bpobj, preferably the empty dummy one (dp_empty_bpobj).
36  */
37 uint64_t
38 bpobj_alloc_empty(objset_t *os, int blocksize, dmu_tx_t *tx)
39 {
40 	spa_t *spa = dmu_objset_spa(os);
41 	dsl_pool_t *dp = dmu_objset_pool(os);
42 
43 	if (spa_feature_is_enabled(spa, SPA_FEATURE_EMPTY_BPOBJ)) {
44 		if (!spa_feature_is_active(spa, SPA_FEATURE_EMPTY_BPOBJ)) {
45 			ASSERT0(dp->dp_empty_bpobj);
46 			dp->dp_empty_bpobj =
47 			    bpobj_alloc(os, SPA_OLD_MAXBLOCKSIZE, tx);
48 			VERIFY(zap_add(os,
49 			    DMU_POOL_DIRECTORY_OBJECT,
50 			    DMU_POOL_EMPTY_BPOBJ, sizeof (uint64_t), 1,
51 			    &dp->dp_empty_bpobj, tx) == 0);
52 		}
53 		spa_feature_incr(spa, SPA_FEATURE_EMPTY_BPOBJ, tx);
54 		ASSERT(dp->dp_empty_bpobj != 0);
55 		return (dp->dp_empty_bpobj);
56 	} else {
57 		return (bpobj_alloc(os, blocksize, tx));
58 	}
59 }
60 
61 void
62 bpobj_decr_empty(objset_t *os, dmu_tx_t *tx)
63 {
64 	dsl_pool_t *dp = dmu_objset_pool(os);
65 
66 	spa_feature_decr(dmu_objset_spa(os), SPA_FEATURE_EMPTY_BPOBJ, tx);
67 	if (!spa_feature_is_active(dmu_objset_spa(os),
68 	    SPA_FEATURE_EMPTY_BPOBJ)) {
69 		VERIFY3U(0, ==, zap_remove(dp->dp_meta_objset,
70 		    DMU_POOL_DIRECTORY_OBJECT,
71 		    DMU_POOL_EMPTY_BPOBJ, tx));
72 		VERIFY3U(0, ==, dmu_object_free(os, dp->dp_empty_bpobj, tx));
73 		dp->dp_empty_bpobj = 0;
74 	}
75 }
76 
77 uint64_t
78 bpobj_alloc(objset_t *os, int blocksize, dmu_tx_t *tx)
79 {
80 	int size;
81 
82 	if (spa_version(dmu_objset_spa(os)) < SPA_VERSION_BPOBJ_ACCOUNT)
83 		size = BPOBJ_SIZE_V0;
84 	else if (spa_version(dmu_objset_spa(os)) < SPA_VERSION_DEADLISTS)
85 		size = BPOBJ_SIZE_V1;
86 	else if (!spa_feature_is_active(dmu_objset_spa(os),
87 	    SPA_FEATURE_LIVELIST))
88 		size = BPOBJ_SIZE_V2;
89 	else
90 		size = sizeof (bpobj_phys_t);
91 
92 	return (dmu_object_alloc(os, DMU_OT_BPOBJ, blocksize,
93 	    DMU_OT_BPOBJ_HDR, size, tx));
94 }
95 
96 void
97 bpobj_free(objset_t *os, uint64_t obj, dmu_tx_t *tx)
98 {
99 	int64_t i;
100 	bpobj_t bpo;
101 	dmu_object_info_t doi;
102 	int epb;
103 	dmu_buf_t *dbuf = NULL;
104 
105 	ASSERT(obj != dmu_objset_pool(os)->dp_empty_bpobj);
106 	VERIFY3U(0, ==, bpobj_open(&bpo, os, obj));
107 
108 	mutex_enter(&bpo.bpo_lock);
109 
110 	if (!bpo.bpo_havesubobj || bpo.bpo_phys->bpo_subobjs == 0)
111 		goto out;
112 
113 	VERIFY3U(0, ==, dmu_object_info(os, bpo.bpo_phys->bpo_subobjs, &doi));
114 	epb = doi.doi_data_block_size / sizeof (uint64_t);
115 
116 	for (i = bpo.bpo_phys->bpo_num_subobjs - 1; i >= 0; i--) {
117 		uint64_t *objarray;
118 		uint64_t offset, blkoff;
119 
120 		offset = i * sizeof (uint64_t);
121 		blkoff = P2PHASE(i, epb);
122 
123 		if (dbuf == NULL || dbuf->db_offset > offset) {
124 			if (dbuf)
125 				dmu_buf_rele(dbuf, FTAG);
126 			VERIFY3U(0, ==, dmu_buf_hold(os,
127 			    bpo.bpo_phys->bpo_subobjs, offset, FTAG, &dbuf, 0));
128 		}
129 
130 		ASSERT3U(offset, >=, dbuf->db_offset);
131 		ASSERT3U(offset, <, dbuf->db_offset + dbuf->db_size);
132 
133 		objarray = dbuf->db_data;
134 		bpobj_free(os, objarray[blkoff], tx);
135 	}
136 	if (dbuf) {
137 		dmu_buf_rele(dbuf, FTAG);
138 		dbuf = NULL;
139 	}
140 	VERIFY3U(0, ==, dmu_object_free(os, bpo.bpo_phys->bpo_subobjs, tx));
141 
142 out:
143 	mutex_exit(&bpo.bpo_lock);
144 	bpobj_close(&bpo);
145 
146 	VERIFY3U(0, ==, dmu_object_free(os, obj, tx));
147 }
148 
149 int
150 bpobj_open(bpobj_t *bpo, objset_t *os, uint64_t object)
151 {
152 	dmu_object_info_t doi;
153 	int err;
154 
155 	err = dmu_object_info(os, object, &doi);
156 	if (err)
157 		return (err);
158 
159 	memset(bpo, 0, sizeof (*bpo));
160 	mutex_init(&bpo->bpo_lock, NULL, MUTEX_DEFAULT, NULL);
161 
162 	ASSERT(bpo->bpo_dbuf == NULL);
163 	ASSERT(bpo->bpo_phys == NULL);
164 	ASSERT(object != 0);
165 	ASSERT3U(doi.doi_type, ==, DMU_OT_BPOBJ);
166 	ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_BPOBJ_HDR);
167 
168 	err = dmu_bonus_hold(os, object, bpo, &bpo->bpo_dbuf);
169 	if (err)
170 		return (err);
171 
172 	bpo->bpo_os = os;
173 	bpo->bpo_object = object;
174 	bpo->bpo_epb = doi.doi_data_block_size >> SPA_BLKPTRSHIFT;
175 	bpo->bpo_havecomp = (doi.doi_bonus_size > BPOBJ_SIZE_V0);
176 	bpo->bpo_havesubobj = (doi.doi_bonus_size > BPOBJ_SIZE_V1);
177 	bpo->bpo_havefreed = (doi.doi_bonus_size > BPOBJ_SIZE_V2);
178 	bpo->bpo_phys = bpo->bpo_dbuf->db_data;
179 	return (0);
180 }
181 
182 boolean_t
183 bpobj_is_open(const bpobj_t *bpo)
184 {
185 	return (bpo->bpo_object != 0);
186 }
187 
188 void
189 bpobj_close(bpobj_t *bpo)
190 {
191 	/* Lame workaround for closing a bpobj that was never opened. */
192 	if (bpo->bpo_object == 0)
193 		return;
194 
195 	dmu_buf_rele(bpo->bpo_dbuf, bpo);
196 	if (bpo->bpo_cached_dbuf != NULL)
197 		dmu_buf_rele(bpo->bpo_cached_dbuf, bpo);
198 	bpo->bpo_dbuf = NULL;
199 	bpo->bpo_phys = NULL;
200 	bpo->bpo_cached_dbuf = NULL;
201 	bpo->bpo_object = 0;
202 
203 	mutex_destroy(&bpo->bpo_lock);
204 }
205 
206 static boolean_t
207 bpobj_is_empty_impl(bpobj_t *bpo)
208 {
209 	ASSERT(MUTEX_HELD(&bpo->bpo_lock));
210 	return (bpo->bpo_phys->bpo_num_blkptrs == 0 &&
211 	    (!bpo->bpo_havesubobj || bpo->bpo_phys->bpo_num_subobjs == 0));
212 }
213 
214 boolean_t
215 bpobj_is_empty(bpobj_t *bpo)
216 {
217 	mutex_enter(&bpo->bpo_lock);
218 	boolean_t is_empty = bpobj_is_empty_impl(bpo);
219 	mutex_exit(&bpo->bpo_lock);
220 	return (is_empty);
221 }
222 
223 /*
224  * A recursive iteration of the bpobjs would be nice here but we run the risk
225  * of overflowing function stack space.  Instead, find each subobj and add it
226  * to the head of our list so it can be scanned for subjobjs.  Like a
227  * recursive implementation, the "deepest" subobjs will be freed first.
228  * When a subobj is found to have no additional subojs, free it.
229  */
230 typedef struct bpobj_info {
231 	bpobj_t *bpi_bpo;
232 	/*
233 	 * This object is a subobj of bpi_parent,
234 	 * at bpi_index in its subobj array.
235 	 */
236 	struct bpobj_info *bpi_parent;
237 	uint64_t bpi_index;
238 	/* How many of our subobj's are left to process. */
239 	uint64_t bpi_unprocessed_subobjs;
240 	/* True after having visited this bpo's directly referenced BPs. */
241 	boolean_t bpi_visited;
242 	list_node_t bpi_node;
243 } bpobj_info_t;
244 
245 static bpobj_info_t *
246 bpi_alloc(bpobj_t *bpo, bpobj_info_t *parent, uint64_t index)
247 {
248 	bpobj_info_t *bpi = kmem_zalloc(sizeof (bpobj_info_t), KM_SLEEP);
249 	bpi->bpi_bpo = bpo;
250 	bpi->bpi_parent = parent;
251 	bpi->bpi_index = index;
252 	if (bpo->bpo_havesubobj && bpo->bpo_phys->bpo_subobjs != 0) {
253 		bpi->bpi_unprocessed_subobjs = bpo->bpo_phys->bpo_num_subobjs;
254 	}
255 	return (bpi);
256 }
257 
258 /*
259  * Update bpobj and all of its parents with new space accounting.
260  */
261 static void
262 propagate_space_reduction(bpobj_info_t *bpi, int64_t freed,
263     int64_t comp_freed, int64_t uncomp_freed, dmu_tx_t *tx)
264 {
265 
266 	for (; bpi != NULL; bpi = bpi->bpi_parent) {
267 		bpobj_t *p = bpi->bpi_bpo;
268 		ASSERT(dmu_buf_is_dirty(p->bpo_dbuf, tx));
269 		p->bpo_phys->bpo_bytes -= freed;
270 		ASSERT3S(p->bpo_phys->bpo_bytes, >=, 0);
271 		if (p->bpo_havecomp) {
272 			p->bpo_phys->bpo_comp -= comp_freed;
273 			p->bpo_phys->bpo_uncomp -= uncomp_freed;
274 		}
275 	}
276 }
277 
278 static int
279 bpobj_iterate_blkptrs(bpobj_info_t *bpi, bpobj_itor_t func, void *arg,
280     int64_t start, dmu_tx_t *tx, boolean_t free)
281 {
282 	int err = 0;
283 	int64_t freed = 0, comp_freed = 0, uncomp_freed = 0;
284 	dmu_buf_t *dbuf = NULL;
285 	bpobj_t *bpo = bpi->bpi_bpo;
286 
287 	int64_t i = bpo->bpo_phys->bpo_num_blkptrs - 1;
288 	uint64_t pe = P2ALIGN_TYPED(i, bpo->bpo_epb, uint64_t) *
289 	    sizeof (blkptr_t);
290 	uint64_t ps = start * sizeof (blkptr_t);
291 	uint64_t pb = MAX((pe > dmu_prefetch_max) ? pe - dmu_prefetch_max : 0,
292 	    ps);
293 	if (pe > pb) {
294 		dmu_prefetch(bpo->bpo_os, bpo->bpo_object, 0, pb, pe - pb,
295 		    ZIO_PRIORITY_ASYNC_READ);
296 	}
297 	for (; i >= start; i--) {
298 		uint64_t offset = i * sizeof (blkptr_t);
299 		uint64_t blkoff = P2PHASE(i, bpo->bpo_epb);
300 
301 		if (dbuf == NULL || dbuf->db_offset > offset) {
302 			if (dbuf)
303 				dmu_buf_rele(dbuf, FTAG);
304 			err = dmu_buf_hold(bpo->bpo_os, bpo->bpo_object,
305 			    offset, FTAG, &dbuf, DMU_READ_NO_PREFETCH);
306 			if (err)
307 				break;
308 			pe = pb;
309 			pb = MAX((dbuf->db_offset > dmu_prefetch_max) ?
310 			    dbuf->db_offset - dmu_prefetch_max : 0, ps);
311 			if (pe > pb) {
312 				dmu_prefetch(bpo->bpo_os, bpo->bpo_object, 0,
313 				    pb, pe - pb, ZIO_PRIORITY_ASYNC_READ);
314 			}
315 		}
316 
317 		ASSERT3U(offset, >=, dbuf->db_offset);
318 		ASSERT3U(offset, <, dbuf->db_offset + dbuf->db_size);
319 
320 		blkptr_t *bparray = dbuf->db_data;
321 		blkptr_t *bp = &bparray[blkoff];
322 
323 		boolean_t bp_freed = BP_GET_FREE(bp);
324 		err = func(arg, bp, bp_freed, tx);
325 		if (err)
326 			break;
327 
328 		if (free) {
329 			int sign = bp_freed ? -1 : +1;
330 			spa_t *spa = dmu_objset_spa(bpo->bpo_os);
331 			freed += sign * bp_get_dsize_sync(spa, bp);
332 			comp_freed += sign * BP_GET_PSIZE(bp);
333 			uncomp_freed += sign * BP_GET_UCSIZE(bp);
334 			ASSERT(dmu_buf_is_dirty(bpo->bpo_dbuf, tx));
335 			bpo->bpo_phys->bpo_num_blkptrs--;
336 			ASSERT3S(bpo->bpo_phys->bpo_num_blkptrs, >=, 0);
337 			if (bp_freed) {
338 				ASSERT(bpo->bpo_havefreed);
339 				bpo->bpo_phys->bpo_num_freed--;
340 				ASSERT3S(bpo->bpo_phys->bpo_num_freed, >=, 0);
341 			}
342 		}
343 	}
344 	if (free) {
345 		propagate_space_reduction(bpi, freed, comp_freed,
346 		    uncomp_freed, tx);
347 		VERIFY0(dmu_free_range(bpo->bpo_os,
348 		    bpo->bpo_object,
349 		    bpo->bpo_phys->bpo_num_blkptrs * sizeof (blkptr_t),
350 		    DMU_OBJECT_END, tx));
351 	}
352 	if (dbuf) {
353 		dmu_buf_rele(dbuf, FTAG);
354 		dbuf = NULL;
355 	}
356 	return (err);
357 }
358 
359 /*
360  * Given an initial bpo, start by freeing the BPs that are directly referenced
361  * by that bpo. If the bpo has subobjs, read in its last subobj and push the
362  * subobj to our stack. By popping items off our stack, eventually we will
363  * encounter a bpo that has no subobjs.  We can free its bpobj_info_t, and if
364  * requested also free the now-empty bpo from disk and decrement
365  * its parent's subobj count. We continue popping each subobj from our stack,
366  * visiting its last subobj until they too have no more subobjs, and so on.
367  */
368 static int
369 bpobj_iterate_impl(bpobj_t *initial_bpo, bpobj_itor_t func, void *arg,
370     dmu_tx_t *tx, boolean_t free, uint64_t *bpobj_size)
371 {
372 	list_t stack;
373 	bpobj_info_t *bpi;
374 	int err = 0;
375 
376 	/*
377 	 * Create a "stack" for us to work with without worrying about
378 	 * stack overflows. Initialize it with the initial_bpo.
379 	 */
380 	list_create(&stack, sizeof (bpobj_info_t),
381 	    offsetof(bpobj_info_t, bpi_node));
382 	mutex_enter(&initial_bpo->bpo_lock);
383 
384 	if (bpobj_size != NULL)
385 		*bpobj_size = initial_bpo->bpo_phys->bpo_num_blkptrs;
386 
387 	list_insert_head(&stack, bpi_alloc(initial_bpo, NULL, 0));
388 
389 	while ((bpi = list_head(&stack)) != NULL) {
390 		bpobj_t *bpo = bpi->bpi_bpo;
391 
392 		ASSERT3P(bpo, !=, NULL);
393 		ASSERT(MUTEX_HELD(&bpo->bpo_lock));
394 		ASSERT(bpobj_is_open(bpo));
395 
396 		if (free)
397 			dmu_buf_will_dirty(bpo->bpo_dbuf, tx);
398 
399 		if (bpi->bpi_visited == B_FALSE) {
400 			err = bpobj_iterate_blkptrs(bpi, func, arg, 0, tx,
401 			    free);
402 			bpi->bpi_visited = B_TRUE;
403 			if (err != 0)
404 				break;
405 		}
406 		/*
407 		 * We've finished with this bpo's directly-referenced BP's and
408 		 * it has no more unprocessed subobjs. We can free its
409 		 * bpobj_info_t (unless it is the topmost, initial_bpo).
410 		 * If we are freeing from disk, we can also do that.
411 		 */
412 		if (bpi->bpi_unprocessed_subobjs == 0) {
413 			/*
414 			 * If there are no entries, there should
415 			 * be no bytes.
416 			 */
417 			if (bpobj_is_empty_impl(bpo)) {
418 				ASSERT0(bpo->bpo_phys->bpo_bytes);
419 				ASSERT0(bpo->bpo_phys->bpo_comp);
420 				ASSERT0(bpo->bpo_phys->bpo_uncomp);
421 			}
422 
423 			/* The initial_bpo has no parent and is not closed. */
424 			if (bpi->bpi_parent != NULL) {
425 				if (free) {
426 					bpobj_t *p = bpi->bpi_parent->bpi_bpo;
427 
428 					ASSERT0(bpo->bpo_phys->bpo_num_blkptrs);
429 					ASSERT3U(p->bpo_phys->bpo_num_subobjs,
430 					    >, 0);
431 					ASSERT3U(bpi->bpi_index, ==,
432 					    p->bpo_phys->bpo_num_subobjs - 1);
433 					ASSERT(dmu_buf_is_dirty(bpo->bpo_dbuf,
434 					    tx));
435 
436 					p->bpo_phys->bpo_num_subobjs--;
437 
438 					VERIFY0(dmu_free_range(p->bpo_os,
439 					    p->bpo_phys->bpo_subobjs,
440 					    bpi->bpi_index * sizeof (uint64_t),
441 					    sizeof (uint64_t), tx));
442 
443 					/* eliminate the empty subobj list */
444 					if (bpo->bpo_havesubobj &&
445 					    bpo->bpo_phys->bpo_subobjs != 0) {
446 						ASSERT0(bpo->bpo_phys->
447 						    bpo_num_subobjs);
448 						err = dmu_object_free(
449 						    bpo->bpo_os,
450 						    bpo->bpo_phys->bpo_subobjs,
451 						    tx);
452 						if (err)
453 							break;
454 						bpo->bpo_phys->bpo_subobjs = 0;
455 					}
456 					err = dmu_object_free(p->bpo_os,
457 					    bpo->bpo_object, tx);
458 					if (err)
459 						break;
460 				}
461 
462 				mutex_exit(&bpo->bpo_lock);
463 				bpobj_close(bpo);
464 				kmem_free(bpo, sizeof (bpobj_t));
465 			} else {
466 				mutex_exit(&bpo->bpo_lock);
467 			}
468 
469 			/*
470 			 * Finished processing this bpo. Unlock, and free
471 			 * our "stack" info.
472 			 */
473 			list_remove_head(&stack);
474 			kmem_free(bpi, sizeof (bpobj_info_t));
475 		} else {
476 			/*
477 			 * We have unprocessed subobjs. Process the next one.
478 			 */
479 			ASSERT(bpo->bpo_havecomp);
480 			ASSERT3P(bpobj_size, ==, NULL);
481 
482 			/* Add the last subobj to stack. */
483 			int64_t i = bpi->bpi_unprocessed_subobjs - 1;
484 			uint64_t offset = i * sizeof (uint64_t);
485 
486 			uint64_t subobj;
487 			err = dmu_read(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs,
488 			    offset, sizeof (uint64_t), &subobj,
489 			    DMU_READ_NO_PREFETCH);
490 			if (err)
491 				break;
492 
493 			bpobj_t *subbpo = kmem_alloc(sizeof (bpobj_t),
494 			    KM_SLEEP);
495 			err = bpobj_open(subbpo, bpo->bpo_os, subobj);
496 			if (err) {
497 				kmem_free(subbpo, sizeof (bpobj_t));
498 				break;
499 			}
500 
501 			if (subbpo->bpo_havesubobj &&
502 			    subbpo->bpo_phys->bpo_subobjs != 0) {
503 				dmu_prefetch(subbpo->bpo_os,
504 				    subbpo->bpo_phys->bpo_subobjs, 0, 0, 0,
505 				    ZIO_PRIORITY_ASYNC_READ);
506 			}
507 
508 			list_insert_head(&stack, bpi_alloc(subbpo, bpi, i));
509 			mutex_enter(&subbpo->bpo_lock);
510 			bpi->bpi_unprocessed_subobjs--;
511 		}
512 	}
513 	/*
514 	 * Cleanup anything left on the "stack" after we left the loop.
515 	 * Every bpo on the stack is locked so we must remember to undo
516 	 * that now (in LIFO order).
517 	 */
518 	while ((bpi = list_remove_head(&stack)) != NULL) {
519 		bpobj_t *bpo = bpi->bpi_bpo;
520 		ASSERT(err != 0);
521 		ASSERT3P(bpo, !=, NULL);
522 
523 		mutex_exit(&bpo->bpo_lock);
524 
525 		/* do not free the initial_bpo */
526 		if (bpi->bpi_parent != NULL) {
527 			bpobj_close(bpi->bpi_bpo);
528 			kmem_free(bpi->bpi_bpo, sizeof (bpobj_t));
529 		}
530 		kmem_free(bpi, sizeof (bpobj_info_t));
531 	}
532 
533 	list_destroy(&stack);
534 
535 	return (err);
536 }
537 
538 /*
539  * Iterate and remove the entries.  If func returns nonzero, iteration
540  * will stop and that entry will not be removed.
541  */
542 int
543 bpobj_iterate(bpobj_t *bpo, bpobj_itor_t func, void *arg, dmu_tx_t *tx)
544 {
545 	return (bpobj_iterate_impl(bpo, func, arg, tx, B_TRUE, NULL));
546 }
547 
548 /*
549  * Iterate the entries.  If func returns nonzero, iteration will stop.
550  *
551  * If there are no subobjs:
552  *
553  * *bpobj_size can be used to return the number of block pointers in the
554  * bpobj.  Note that this may be different from the number of block pointers
555  * that are iterated over, if iteration is terminated early (e.g. by the func
556  * returning nonzero).
557  *
558  * If there are concurrent (or subsequent) modifications to the bpobj then the
559  * returned *bpobj_size can be passed as "start" to
560  * livelist_bpobj_iterate_from_nofree() to iterate the newly added entries.
561  */
562 int
563 bpobj_iterate_nofree(bpobj_t *bpo, bpobj_itor_t func, void *arg,
564     uint64_t *bpobj_size)
565 {
566 	return (bpobj_iterate_impl(bpo, func, arg, NULL, B_FALSE, bpobj_size));
567 }
568 
569 /*
570  * Iterate over the blkptrs in the bpobj beginning at index start. If func
571  * returns nonzero, iteration will stop. This is a livelist specific function
572  * since it assumes that there are no subobjs present.
573  */
574 int
575 livelist_bpobj_iterate_from_nofree(bpobj_t *bpo, bpobj_itor_t func, void *arg,
576     int64_t start)
577 {
578 	if (bpo->bpo_havesubobj)
579 		VERIFY0(bpo->bpo_phys->bpo_subobjs);
580 	bpobj_info_t *bpi = bpi_alloc(bpo, NULL, 0);
581 	int err = bpobj_iterate_blkptrs(bpi, func, arg, start, NULL, B_FALSE);
582 	kmem_free(bpi, sizeof (bpobj_info_t));
583 	return (err);
584 }
585 
586 /*
587  * Logically add subobj's contents to the parent bpobj.
588  *
589  * In the most general case, this is accomplished in constant time by adding
590  * a reference to subobj.  This case is used when enqueuing a large subobj:
591  * +--------------+                        +--------------+
592  * | bpobj        |----------------------->| subobj list  |
593  * +----+----+----+----+----+              +-----+-----+--+--+
594  * | bp | bp | bp | bp | bp |              | obj | obj | obj |
595  * +----+----+----+----+----+              +-----+-----+-----+
596  *
597  * +--------------+                        +--------------+
598  * | sub-bpobj    |----------------------> | subsubobj    |
599  * +----+----+----+----+---------+----+    +-----+-----+--+--------+-----+
600  * | bp | bp | bp | bp |   ...   | bp |    | obj | obj |    ...    | obj |
601  * +----+----+----+----+---------+----+    +-----+-----+-----------+-----+
602  *
603  * Result: sub-bpobj added to parent's subobj list.
604  * +--------------+                        +--------------+
605  * | bpobj        |----------------------->| subobj list  |
606  * +----+----+----+----+----+              +-----+-----+--+--+-----+
607  * | bp | bp | bp | bp | bp |              | obj | obj | obj | OBJ |
608  * +----+----+----+----+----+              +-----+-----+-----+--|--+
609  *                                                              |
610  *       /-----------------------------------------------------/
611  *       v
612  * +--------------+                        +--------------+
613  * | sub-bpobj    |----------------------> | subsubobj    |
614  * +----+----+----+----+---------+----+    +-----+-----+--+--------+-----+
615  * | bp | bp | bp | bp |   ...   | bp |    | obj | obj |    ...    | obj |
616  * +----+----+----+----+---------+----+    +-----+-----+-----------+-----+
617  *
618  *
619  * In a common case, the subobj is small: its bp's and its list of subobj's
620  * are each stored in a single block.  In this case we copy the subobj's
621  * contents to the parent:
622  * +--------------+                        +--------------+
623  * | bpobj        |----------------------->| subobj list  |
624  * +----+----+----+----+----+              +-----+-----+--+--+
625  * | bp | bp | bp | bp | bp |              | obj | obj | obj |
626  * +----+----+----+----+----+              +-----+-----+-----+
627  *                          ^                                ^
628  * +--------------+         |              +--------------+  |
629  * | sub-bpobj    |---------^------------> | subsubobj    |  ^
630  * +----+----+----+         |              +-----+-----+--+  |
631  * | BP | BP |-->-->-->-->-/               | OBJ | OBJ |-->-/
632  * +----+----+                             +-----+-----+
633  *
634  * Result: subobj destroyed, contents copied to parent:
635  * +--------------+                        +--------------+
636  * | bpobj        |----------------------->| subobj list  |
637  * +----+----+----+----+----+----+----+    +-----+-----+--+--+-----+-----+
638  * | bp | bp | bp | bp | bp | BP | BP |    | obj | obj | obj | OBJ | OBJ |
639  * +----+----+----+----+----+----+----+    +-----+-----+-----+-----+-----+
640  *
641  *
642  * If the subobj has many BP's but few subobj's, we can copy the sub-subobj's
643  * but retain the sub-bpobj:
644  * +--------------+                        +--------------+
645  * | bpobj        |----------------------->| subobj list  |
646  * +----+----+----+----+----+              +-----+-----+--+--+
647  * | bp | bp | bp | bp | bp |              | obj | obj | obj |
648  * +----+----+----+----+----+              +-----+-----+-----+
649  *                                                           ^
650  * +--------------+                        +--------------+  |
651  * | sub-bpobj    |----------------------> | subsubobj    |  ^
652  * +----+----+----+----+---------+----+    +-----+-----+--+  |
653  * | bp | bp | bp | bp |   ...   | bp |    | OBJ | OBJ |-->-/
654  * +----+----+----+----+---------+----+    +-----+-----+
655  *
656  * Result: sub-sub-bpobjs and subobj added to parent's subobj list.
657  * +--------------+                     +--------------+
658  * | bpobj        |-------------------->| subobj list  |
659  * +----+----+----+----+----+           +-----+-----+--+--+-----+-----+------+
660  * | bp | bp | bp | bp | bp |           | obj | obj | obj | OBJ | OBJ | OBJ* |
661  * +----+----+----+----+----+           +-----+-----+-----+-----+-----+--|---+
662  *                                                                       |
663  *       /--------------------------------------------------------------/
664  *       v
665  * +--------------+
666  * | sub-bpobj    |
667  * +----+----+----+----+---------+----+
668  * | bp | bp | bp | bp |   ...   | bp |
669  * +----+----+----+----+---------+----+
670  */
671 void
672 bpobj_enqueue_subobj(bpobj_t *bpo, uint64_t subobj, dmu_tx_t *tx)
673 {
674 	bpobj_t subbpo;
675 	uint64_t used, comp, uncomp, subsubobjs;
676 	boolean_t copy_subsub = B_TRUE;
677 	boolean_t copy_bps = B_TRUE;
678 
679 	ASSERT(bpobj_is_open(bpo));
680 	ASSERT(subobj != 0);
681 	ASSERT(bpo->bpo_havesubobj);
682 	ASSERT(bpo->bpo_havecomp);
683 	ASSERT(bpo->bpo_object != dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj);
684 
685 	if (subobj == dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj) {
686 		bpobj_decr_empty(bpo->bpo_os, tx);
687 		return;
688 	}
689 
690 	VERIFY3U(0, ==, bpobj_open(&subbpo, bpo->bpo_os, subobj));
691 	if (bpobj_is_empty(&subbpo)) {
692 		/* No point in having an empty subobj. */
693 		bpobj_close(&subbpo);
694 		bpobj_free(bpo->bpo_os, subobj, tx);
695 		return;
696 	}
697 	VERIFY3U(0, ==, bpobj_space(&subbpo, &used, &comp, &uncomp));
698 
699 	mutex_enter(&bpo->bpo_lock);
700 	dmu_buf_will_dirty(bpo->bpo_dbuf, tx);
701 
702 	dmu_object_info_t doi;
703 
704 	if (bpo->bpo_phys->bpo_subobjs != 0) {
705 		ASSERT0(dmu_object_info(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs,
706 		    &doi));
707 		ASSERT3U(doi.doi_type, ==, DMU_OT_BPOBJ_SUBOBJ);
708 	}
709 
710 	/*
711 	 * If subobj has only one block of subobjs, then move subobj's
712 	 * subobjs to bpo's subobj list directly.  This reduces recursion in
713 	 * bpobj_iterate due to nested subobjs.
714 	 */
715 	subsubobjs = subbpo.bpo_phys->bpo_subobjs;
716 	if (subsubobjs != 0) {
717 		VERIFY0(dmu_object_info(bpo->bpo_os, subsubobjs, &doi));
718 		if (doi.doi_max_offset > doi.doi_data_block_size) {
719 			copy_subsub = B_FALSE;
720 		}
721 	}
722 
723 	/*
724 	 * If, in addition to having only one block of subobj's, subobj has
725 	 * only one block of bp's, then move subobj's bp's to bpo's bp list
726 	 * directly. This reduces recursion in bpobj_iterate due to nested
727 	 * subobjs.
728 	 */
729 	VERIFY3U(0, ==, dmu_object_info(bpo->bpo_os, subobj, &doi));
730 	if (doi.doi_max_offset > doi.doi_data_block_size || !copy_subsub) {
731 		copy_bps = B_FALSE;
732 	}
733 
734 	if (copy_subsub && subsubobjs != 0) {
735 		dmu_buf_t *subdb;
736 		uint64_t numsubsub = subbpo.bpo_phys->bpo_num_subobjs;
737 
738 		VERIFY0(dmu_buf_hold(bpo->bpo_os, subsubobjs,
739 		    0, FTAG, &subdb, 0));
740 		/*
741 		 * Make sure that we are not asking dmu_write()
742 		 * to write more data than we have in our buffer.
743 		 */
744 		VERIFY3U(subdb->db_size, >=,
745 		    numsubsub * sizeof (subobj));
746 		if (bpo->bpo_phys->bpo_subobjs == 0) {
747 			bpo->bpo_phys->bpo_subobjs =
748 			    dmu_object_alloc(bpo->bpo_os,
749 			    DMU_OT_BPOBJ_SUBOBJ, SPA_OLD_MAXBLOCKSIZE,
750 			    DMU_OT_NONE, 0, tx);
751 		}
752 		dmu_write(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs,
753 		    bpo->bpo_phys->bpo_num_subobjs * sizeof (subobj),
754 		    numsubsub * sizeof (subobj), subdb->db_data, tx);
755 		dmu_buf_rele(subdb, FTAG);
756 		bpo->bpo_phys->bpo_num_subobjs += numsubsub;
757 
758 		dmu_buf_will_dirty(subbpo.bpo_dbuf, tx);
759 		subbpo.bpo_phys->bpo_subobjs = 0;
760 		VERIFY0(dmu_object_free(bpo->bpo_os, subsubobjs, tx));
761 	}
762 
763 	if (copy_bps) {
764 		dmu_buf_t *bps;
765 		uint64_t numbps = subbpo.bpo_phys->bpo_num_blkptrs;
766 
767 		ASSERT(copy_subsub);
768 		VERIFY0(dmu_buf_hold(bpo->bpo_os, subobj,
769 		    0, FTAG, &bps, 0));
770 
771 		/*
772 		 * Make sure that we are not asking dmu_write()
773 		 * to write more data than we have in our buffer.
774 		 */
775 		VERIFY3U(bps->db_size, >=, numbps * sizeof (blkptr_t));
776 		dmu_write(bpo->bpo_os, bpo->bpo_object,
777 		    bpo->bpo_phys->bpo_num_blkptrs * sizeof (blkptr_t),
778 		    numbps * sizeof (blkptr_t),
779 		    bps->db_data, tx);
780 		dmu_buf_rele(bps, FTAG);
781 		bpo->bpo_phys->bpo_num_blkptrs += numbps;
782 
783 		bpobj_close(&subbpo);
784 		VERIFY0(dmu_object_free(bpo->bpo_os, subobj, tx));
785 	} else {
786 		bpobj_close(&subbpo);
787 		if (bpo->bpo_phys->bpo_subobjs == 0) {
788 			bpo->bpo_phys->bpo_subobjs =
789 			    dmu_object_alloc(bpo->bpo_os,
790 			    DMU_OT_BPOBJ_SUBOBJ, SPA_OLD_MAXBLOCKSIZE,
791 			    DMU_OT_NONE, 0, tx);
792 		}
793 
794 		dmu_write(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs,
795 		    bpo->bpo_phys->bpo_num_subobjs * sizeof (subobj),
796 		    sizeof (subobj), &subobj, tx);
797 		bpo->bpo_phys->bpo_num_subobjs++;
798 	}
799 
800 	bpo->bpo_phys->bpo_bytes += used;
801 	bpo->bpo_phys->bpo_comp += comp;
802 	bpo->bpo_phys->bpo_uncomp += uncomp;
803 	mutex_exit(&bpo->bpo_lock);
804 
805 }
806 
807 /*
808  * Prefetch metadata required for bpobj_enqueue_subobj().
809  */
810 void
811 bpobj_prefetch_subobj(bpobj_t *bpo, uint64_t subobj)
812 {
813 	dmu_object_info_t doi;
814 	bpobj_t subbpo;
815 	uint64_t subsubobjs;
816 	boolean_t copy_subsub = B_TRUE;
817 	boolean_t copy_bps = B_TRUE;
818 
819 	ASSERT(bpobj_is_open(bpo));
820 	ASSERT(subobj != 0);
821 
822 	if (subobj == dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj)
823 		return;
824 
825 	if (bpobj_open(&subbpo, bpo->bpo_os, subobj) != 0)
826 		return;
827 	if (bpobj_is_empty(&subbpo)) {
828 		bpobj_close(&subbpo);
829 		return;
830 	}
831 	subsubobjs = subbpo.bpo_phys->bpo_subobjs;
832 	bpobj_close(&subbpo);
833 
834 	if (subsubobjs != 0) {
835 		if (dmu_object_info(bpo->bpo_os, subsubobjs, &doi) != 0)
836 			return;
837 		if (doi.doi_max_offset > doi.doi_data_block_size)
838 			copy_subsub = B_FALSE;
839 	}
840 
841 	if (dmu_object_info(bpo->bpo_os, subobj, &doi) != 0)
842 		return;
843 	if (doi.doi_max_offset > doi.doi_data_block_size || !copy_subsub)
844 		copy_bps = B_FALSE;
845 
846 	if (copy_subsub && subsubobjs != 0) {
847 		if (bpo->bpo_phys->bpo_subobjs) {
848 			dmu_prefetch(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs, 0,
849 			    bpo->bpo_phys->bpo_num_subobjs * sizeof (subobj), 1,
850 			    ZIO_PRIORITY_ASYNC_READ);
851 		}
852 		dmu_prefetch(bpo->bpo_os, subsubobjs, 0, 0, 1,
853 		    ZIO_PRIORITY_ASYNC_READ);
854 	}
855 
856 	if (copy_bps) {
857 		dmu_prefetch(bpo->bpo_os, bpo->bpo_object, 0,
858 		    bpo->bpo_phys->bpo_num_blkptrs * sizeof (blkptr_t), 1,
859 		    ZIO_PRIORITY_ASYNC_READ);
860 		dmu_prefetch(bpo->bpo_os, subobj, 0, 0, 1,
861 		    ZIO_PRIORITY_ASYNC_READ);
862 	} else if (bpo->bpo_phys->bpo_subobjs) {
863 		dmu_prefetch(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs, 0,
864 		    bpo->bpo_phys->bpo_num_subobjs * sizeof (subobj), 1,
865 		    ZIO_PRIORITY_ASYNC_READ);
866 	}
867 }
868 
869 void
870 bpobj_enqueue(bpobj_t *bpo, const blkptr_t *bp, boolean_t bp_freed,
871     dmu_tx_t *tx)
872 {
873 	blkptr_t stored_bp = *bp;
874 	uint64_t offset;
875 	int blkoff;
876 	blkptr_t *bparray;
877 
878 	ASSERT(bpobj_is_open(bpo));
879 	ASSERT(!BP_IS_HOLE(bp));
880 	ASSERT(bpo->bpo_object != dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj);
881 
882 	if (BP_IS_EMBEDDED(bp)) {
883 		/*
884 		 * The bpobj will compress better without the payload.
885 		 *
886 		 * Note that we store EMBEDDED bp's because they have an
887 		 * uncompressed size, which must be accounted for.  An
888 		 * alternative would be to add their size to bpo_uncomp
889 		 * without storing the bp, but that would create additional
890 		 * complications: bpo_uncomp would be inconsistent with the
891 		 * set of BP's stored, and bpobj_iterate() wouldn't visit
892 		 * all the space accounted for in the bpobj.
893 		 */
894 		memset(&stored_bp, 0, sizeof (stored_bp));
895 		stored_bp.blk_prop = bp->blk_prop;
896 		BP_SET_LOGICAL_BIRTH(&stored_bp, BP_GET_LOGICAL_BIRTH(bp));
897 	} else if (!BP_GET_DEDUP(bp)) {
898 		/* The bpobj will compress better without the checksum */
899 		memset(&stored_bp.blk_cksum, 0, sizeof (stored_bp.blk_cksum));
900 	}
901 
902 	stored_bp.blk_fill = 0;
903 	BP_SET_FREE(&stored_bp, bp_freed);
904 
905 	mutex_enter(&bpo->bpo_lock);
906 
907 	offset = bpo->bpo_phys->bpo_num_blkptrs * sizeof (stored_bp);
908 	blkoff = P2PHASE(bpo->bpo_phys->bpo_num_blkptrs, bpo->bpo_epb);
909 
910 	if (bpo->bpo_cached_dbuf == NULL ||
911 	    offset < bpo->bpo_cached_dbuf->db_offset ||
912 	    offset >= bpo->bpo_cached_dbuf->db_offset +
913 	    bpo->bpo_cached_dbuf->db_size) {
914 		if (bpo->bpo_cached_dbuf)
915 			dmu_buf_rele(bpo->bpo_cached_dbuf, bpo);
916 		VERIFY3U(0, ==, dmu_buf_hold(bpo->bpo_os, bpo->bpo_object,
917 		    offset, bpo, &bpo->bpo_cached_dbuf, 0));
918 		ASSERT3P(bpo->bpo_cached_dbuf, !=, NULL);
919 	}
920 
921 	dmu_buf_will_dirty(bpo->bpo_cached_dbuf, tx);
922 	bparray = bpo->bpo_cached_dbuf->db_data;
923 	bparray[blkoff] = stored_bp;
924 
925 	dmu_buf_will_dirty(bpo->bpo_dbuf, tx);
926 	bpo->bpo_phys->bpo_num_blkptrs++;
927 	int sign = bp_freed ? -1 : +1;
928 	bpo->bpo_phys->bpo_bytes += sign *
929 	    bp_get_dsize_sync(dmu_objset_spa(bpo->bpo_os), bp);
930 	if (bpo->bpo_havecomp) {
931 		bpo->bpo_phys->bpo_comp += sign * BP_GET_PSIZE(bp);
932 		bpo->bpo_phys->bpo_uncomp += sign * BP_GET_UCSIZE(bp);
933 	}
934 	if (bp_freed) {
935 		ASSERT(bpo->bpo_havefreed);
936 		bpo->bpo_phys->bpo_num_freed++;
937 	}
938 	mutex_exit(&bpo->bpo_lock);
939 }
940 
941 struct space_range_arg {
942 	spa_t *spa;
943 	uint64_t mintxg;
944 	uint64_t maxtxg;
945 	uint64_t used;
946 	uint64_t comp;
947 	uint64_t uncomp;
948 };
949 
950 static int
951 space_range_cb(void *arg, const blkptr_t *bp, boolean_t bp_freed, dmu_tx_t *tx)
952 {
953 	(void) bp_freed, (void) tx;
954 	struct space_range_arg *sra = arg;
955 
956 	if (BP_GET_LOGICAL_BIRTH(bp) > sra->mintxg &&
957 	    BP_GET_LOGICAL_BIRTH(bp) <= sra->maxtxg) {
958 		if (dsl_pool_sync_context(spa_get_dsl(sra->spa)))
959 			sra->used += bp_get_dsize_sync(sra->spa, bp);
960 		else
961 			sra->used += bp_get_dsize(sra->spa, bp);
962 		sra->comp += BP_GET_PSIZE(bp);
963 		sra->uncomp += BP_GET_UCSIZE(bp);
964 	}
965 	return (0);
966 }
967 
968 int
969 bpobj_space(bpobj_t *bpo, uint64_t *usedp, uint64_t *compp, uint64_t *uncompp)
970 {
971 	ASSERT(bpobj_is_open(bpo));
972 	mutex_enter(&bpo->bpo_lock);
973 
974 	*usedp = bpo->bpo_phys->bpo_bytes;
975 	if (bpo->bpo_havecomp) {
976 		*compp = bpo->bpo_phys->bpo_comp;
977 		*uncompp = bpo->bpo_phys->bpo_uncomp;
978 		mutex_exit(&bpo->bpo_lock);
979 		return (0);
980 	} else {
981 		mutex_exit(&bpo->bpo_lock);
982 		return (bpobj_space_range(bpo, 0, UINT64_MAX,
983 		    usedp, compp, uncompp));
984 	}
985 }
986 
987 /*
988  * Return the amount of space in the bpobj which is:
989  * mintxg < logical birth <= maxtxg
990  */
991 int
992 bpobj_space_range(bpobj_t *bpo, uint64_t mintxg, uint64_t maxtxg,
993     uint64_t *usedp, uint64_t *compp, uint64_t *uncompp)
994 {
995 	struct space_range_arg sra = { 0 };
996 	int err;
997 
998 	ASSERT(bpobj_is_open(bpo));
999 
1000 	/*
1001 	 * As an optimization, if they want the whole txg range, just
1002 	 * get bpo_bytes rather than iterating over the bps.
1003 	 */
1004 	if (mintxg < TXG_INITIAL && maxtxg == UINT64_MAX && bpo->bpo_havecomp)
1005 		return (bpobj_space(bpo, usedp, compp, uncompp));
1006 
1007 	sra.spa = dmu_objset_spa(bpo->bpo_os);
1008 	sra.mintxg = mintxg;
1009 	sra.maxtxg = maxtxg;
1010 
1011 	err = bpobj_iterate_nofree(bpo, space_range_cb, &sra, NULL);
1012 	*usedp = sra.used;
1013 	*compp = sra.comp;
1014 	*uncompp = sra.uncomp;
1015 	return (err);
1016 }
1017 
1018 /*
1019  * A bpobj_itor_t to append blkptrs to a bplist. Note that while blkptrs in a
1020  * bpobj are designated as free or allocated that information is not preserved
1021  * in bplists.
1022  */
1023 int
1024 bplist_append_cb(void *arg, const blkptr_t *bp, boolean_t bp_freed,
1025     dmu_tx_t *tx)
1026 {
1027 	(void) bp_freed, (void) tx;
1028 	bplist_t *bpl = arg;
1029 	bplist_append(bpl, bp);
1030 	return (0);
1031 }
1032